Integrating AI into your organization’s workflow and core values can provide a competitive advantage if used responsibly and inclusively.
While generative AI offers enormous potential in the workplace, including increased efficiency and productivity, it also presents significant challenges around gender and racial biases. As organizations are increasingly adopting AI, addressing these biases is more urgent than ever.
The Catalyst webinar “How to Use Generative AI Free of Gender and Racial Bias” explored strategies for creating ethical AI practices that promote equity and inclusion in the workplace. By understanding and mitigating its biases, organizations can leverage AI as a tool for progress rather than one that reinforces inequalities.
Moderated by Julie Cafley, Executive Director, Catalyst Canada, the panel included Cathy Cobey, Global Responsible AI Co-Lead, EY; Noelle Russell, Chief AI Officer, AI Leadership Institute; and Michael Thomson, Executive Vice President, Edelman.
Here are our top five insights from the discussion:
- AI education is essential to avoid pitfalls.
AI can enhance human capabilities, yet it also carries inherent risks. Russell likened AI to magic, highlighting its capacity to leverage past data and behaviors to refine decision making, while Cobey noted that “AI is still human controlled. Maintaining this oversight is essential in AI development. There’s always a magician behind every magic trick.”All panelists agreed that users need better education to utilize AI effectively and avoid potential risks. “There is a dark side to this magic,” Thomson warned. “AI’s transformative power offers great opportunities but requires robust risk management strategies to protect against its pitfalls.”
- Understand that all AI is inherently biased.
AI systems can unintentionally absorb biases from their training data. “No dataset or person is free of bias,” Thomson said.Data scientists aim to create holistic, empathetic, and inclusive models, believing that their innovations will deliver beneficial outcomes. “At the onset, these models are like baby tigers—cute and full of potential,” Russell said. “However, critical questions about their long-term impact are often overlooked, leading to significant challenges as AI models mature and become operational. Ensuring diverse representation throughout the development process is essential to minimize biases effectively.”
- Implement strong AI principles.
AI standards should be ingrained into an organization’s core values, not treated as an afterthought. “AI principles need to be like water in a wave, woven through everything we do,” said Russell. She also advised that leadership must shift their thinking and incorporate these principles into the organization’s core values.Cobey added that EY updated their AI principles in September 2023 to include sustainability, noting the high level of computer processing, energy, and water usage involved in running large language models (LLMs) and other types of AI models. She also pointed out that some of the principles can be in conflict with each other and require trade-offs. “Sometimes security works against transparency and accuracy against explainability,” she said. “You have to choose which principles and values are the most important, depending on the AI use case.”
- Embrace AI—cautiously—for a competitive advantage.
“The best advice I can give is to just start using AI,” Thomson said. “It may be intimidating at first, so start with small, manageable tasks. AI is a tool with limitations. It’s like an imperfect first draft. Don’t blindly share AI-generated information.” He likened it to using a calculator for complex math, saying “Without it, I’d struggle, but with it, I’m efficient. AI can be that tool for you. Organizations using AI effectively will outpace those that aren’t, making AI training essential for everyone. We need to take advantage of these tools, particularly when other people aren’t.” - AI needs all our perspectives for a better future.
Active participation in AI development is the way forward. “We need to help it make better decisions. We are part of the solution,” Cobey said, adding that users don’t need to be tech savvy.Russell stressed the importance of diverse perspectives in building AI systems that serve everyone, saying, “Don’t build for people without those people. It’s important to show, not tell, when advocating for diversity and inclusion.” Outsourcing how AI will be implemented at your organization, she cautioned, is ill-advised without diverse representation.
As organizations navigate the complex challenges of integrating generative AI, they should focus on maintaining ethical standards that promote inclusivity and lessen biases. By conscientiously implementing AI practices that prioritize diversity, equity, and inclusion, organizations will reap the enormous benefits of this transformative technology while continuing to advance workplace equity and support their core values.
Want more insights like these?
Get Catalyst in your inbox and start building a more inclusive future for yourself and your organization.